Gradient Informed, GPU Accelerated Lens modelling (GIGA-Lens) is a package for fast Bayesian inference on strong gravitational lenses. For details, please see our paper. See here for our documentation.
GIGA-Lens can be installed via pip: :
pip install gigalensIf pip notes an error after installation about conflicting dependencies, these can usually be safely ignored. If you wish to test the installation, tests can be run simply by running tox in the root directory.
If you don’t have access to institutional GPUs, one easy way is to use GPU on Google Colab. Please remember the very first cell should have !pip install gigalens. If you do have access to institutional GPUs, you can set up a notebook to run on GPU. For example, at NESRC, you can choose the kernel tensorflow-2.6.0, and include in the first cell: !pip install gigalens.
The following packages are requirements for GIGA-Lens. However, !pip install gigalens is all you need to do. In fact, separately installing other packages can cause issues with subpackage dependencies. Some users may find it necessary to install PyYAML.
tensorflow>=2.6.0
tensorflow-probability>=0.15.0
lenstronomy==1.9.3
scikit-image==0.18.2
tqdm==4.62.0The following dependencies are required by lenstronomy:
cosmohammer==0.6.1
schwimmbad==0.3.2
dynesty==1.1
corner==2.2.1
mpmath==1.2.1GIGALens was written by Andi Gu.